The Role of Design Complexity in Technology Improvement

We study a simplification of a previously proposed model of technology evolution to understand the behavior of performance curves, which describe how a technology improves with increasing cumulative production. The model decomposes a technology or production process into components that get cheaper as new implementations of them are discovered. We show that the model predicts a power-law relation between the cost of a technology and the number of attempts at improving it, in agreement with observation. The model further predicts that the design of a technology may influence both its average rate of improvement and the amount of uncertainty in its improvement. We explore the testability of the model and its implications for measurement of performance curves and for technological improvement in general.